XX

Apodex 1.1 Mini (Free)

apodex/apodex-1.1-mini:free

Apodex 1.1 Mini is Apodex AI’s open-weight, 35B-parameter reasoning model based on Qwen3.5-35B-A3B. It is designed for coding, research and tool-using agents, with a 262,144-token context window and an Apache 2.0 license. Novita offers this free variant with text input and output, reasoning, function calling and structured outputs. Agent orchestration and tool execution require an external harness such as FrontierAgent.

Baichuan-M2 is a medically-enhanced reasoning model specifically designed for real-world medical reasoning tasks. We begin with real-world medical questions and conduct reinforcement learning training based on a large-scale verifier system. While maintaining the model's general capabilities, the medical effectiveness of Baichuan-M2 has achieved breakthrough improvements. Baichuan-M2 is currently the world's best open-source medical model. On the HealthBench Benchmark, it surpasses all open-sour

ByteDance Seed: Seed 2.1 Turbo

bytedance-seed/seed-2-1-turbo

Seed 2.1 Turbo is a multimodal model from ByteDance Seed for coding and long-horizon agent workflows. It is suited for end-to-end software delivery, multi-step task execution, and understanding visual and...

ByteDance Seed: Seed-2.0-Code

bytedance-seed/seed-2.0-code

Seed 2.0 Code is a model from ByteDance Seed optimized for agentic coding. It is suited for frontend development, multilingual programming tasks, and coding-agent workflows in tools such as Claude...

ByteDance Seed: Seed-2.0-Lite

bytedance-seed/seed-2.0-lite

Seed-2.0-Lite is a versatile, cost‑efficient enterprise workhorse that delivers strong multimodal and agent capabilities while offering noticeably lower latency, making it a practical default choice for most production workloads across...

ByteDance Seed: Seed-2.0-Mini

bytedance-seed/seed-2.0-mini

Seed-2.0-mini targets latency-sensitive, high-concurrency, and cost-sensitive scenarios, emphasizing fast response and flexible inference deployment. It delivers performance comparable to ByteDance-Seed-1.6, supports 256k context, four reasoning effort modes (minimal/low/medium/high), multimodal understanding,...

ByteDance Seed: Seed 1.6

bytedance-seed/seed-1.6

Seed 1.6 is a general-purpose model released by the ByteDance Seed team. It incorporates multimodal capabilities and adaptive deep thinking with a 256K context window.

ByteDance Seed: Seed 1.6 Flash

bytedance-seed/seed-1.6-flash

Seed 1.6 Flash is an ultra-fast multimodal deep thinking model by ByteDance Seed, supporting both text and visual understanding. It features a 256k context window and can generate outputs of...

Venice Uncensored Dolphin Mistral 24B Venice Edition is a fine-tuned variant of Mistral-Small-24B-Instruct-2501, developed by dphn.ai in collaboration with Venice.ai. This model is designed as an “uncensored” instruct-tuned LLM, preserving...

DeepSeek V4.1 Flash Uncensored

dealignai/deepseek-v4.1-flash-uncensored

An abliterated FP8 variant of DeepSeek V4.1 Flash by dealign.ai. It accepts text and images and supports reasoning, tool calling and structured outputs, with a one-million-token context window through Venice.

ea-29b-a4b is an open-weight model currently exposed by ModelScope API-Inference.

The idea behind this merge is that each layer is composed of several tensors, which are in turn responsible for specific functions. Using MythoLogic-L2's robust understanding as its input and Huginn's extensive writing capability as its output seems to have resulted in a model that exceeds at both, confirming my theory. (More details to be released at a later time).

Kev-4B is the open-source community edition of Jev, created by Jared Palmer and fine-tuned from Qwen3.5-4B-Base. It returns structured judgments for routing, triage, moderation, and risk control.

Original Kev prototype based on Qwen2.5-0.5B; the author marks it superseded by later Kev checkpoints.

Previous-generation Kev model based on Qwen3-0.6B-Base; superseded by Kev-0.8B.

Kev-0.8B is the current small Kev decision model, based on Qwen3.5-0.8B-Base. It returns calibrated probabilities for typed choice, yes/no, and score questions.

Kev-27B is a current Kev decision model based on post-trained Qwen3.8-27B. It returns calibrated probabilities for typed choice, yes/no, and score questions.

Previous-generation Kev model based on Qwen3-8B-Base; superseded by Kev-9B.

Krea image editing model listed by OpenRouter, accepting up to four reference images and a text instruction.

Krea's large image generation model with style references, moodboards, and creative controls.

Krea's medium image generation model with style references, moodboards, and creative controls.

Krea's medium turbo image generation model with style references, moodboards, and creative controls.

Mancer: Weaver (alpha)

mancer/weaver

An attempt to recreate Claude-style verbosity, but don't expect the same level of coherence or memory. Meant for use in roleplay/narrative situations.

Nous: Hermes 4 405B

nousresearch/hermes-4-405b

Hermes 4 is a large-scale reasoning model built on Meta-Llama-3.1-405B and released by Nous Research. It introduces a hybrid reasoning mode, where the model can choose to deliberate internally with...

Nous: Hermes 4 70B

nousresearch/hermes-4-70b

Hermes 4 70B is a hybrid reasoning model from Nous Research, built on Meta-Llama-3.1-70B. It introduces the same hybrid mode as the larger 405B release, allowing the model to either...

Nous: Hermes 3 70B Instruct

nousresearch/hermes-3-llama-3.1-70b

Hermes 3 is a generalist language model with many improvements over [Hermes 2](/models/nousresearch/nous-hermes-2-mistral-7b-dpo), including advanced agentic capabilities, much better roleplaying, reasoning, multi-turn conversation, long context coherence, and improvements across the...

Nous: Hermes 3 405B Instruct

nousresearch/hermes-3-llama-3.1-405b

Hermes 3 is a generalist language model with many improvements over Hermes 2, including advanced agentic capabilities, much better roleplaying, reasoning, multi-turn conversation, long context coherence, and improvements across the...

Ornith 1.5 35B A3B is an open-weight multimodal model served by RunInfra with text and image input, tool calling, JSON mode, streaming, and a 262,144-token context window.

Beam

reflection/beam-501b-a23b

Reflection’s open-weight, text-only mixture-of-experts model for coding, reasoning and agentic tool use, with 501.3B total and 23.4B active parameters.

Beam (free)

reflection/beam-501b-a23b:free

Reflection’s open-weight, text-only mixture-of-experts model for coding, reasoning and agentic tool use, with 501.3B total and 23.4B active parameters.

Span-02

respan/span-02

Open-weight decision model from Respan for typed choices, yes/no judgments, scores, and numbers from text and images, with probability distributions and anytime reasoning.

Span-01

respan/span-01

Reasoning classifier for behavior probabilities on agent conversation traces. Respan charges $0.02 per million input tokens; output is free. Requires Respan credits and early access.

Span-01 Lite (Free)

respan/span-01:free

Free Lite tier of Respan Span-01 for behavior probabilities on agent conversation traces. Subject to Respan’s daily limit, which resets at 00:00 UTC, and early-access availability.

Agnes 2.5 Flash is Agnes AI's fast multimodal model for coding, reasoning, tool use, multi-turn dialogue, image understanding, and agent workflows, with a 512K-token context window.

Agnes 2.5 Pro is Agnes AI's higher-capability multimodal reasoning model for complex coding, visual understanding, and long-context agent workflows, with a 1M-token context window.

Agnes 2.0 Flash is Agnes AI's multimodal model for coding, reasoning, tool use, agent workflows, and image understanding, with a 256K-token context window.

ScaleDown Classify-1 answers typed Noul, Choice and Score questions with calibrated probabilities instead of generated text.

ScaleDown Extract finds the fields you describe in a JSON schema and returns each match with a confidence score, character offsets and surrounding context, as JSON in the message content.

ScaleDown Summarize writes abstractive summaries of long text. A system message can set format, focus or length, and the summary is returned as JSON in the message content.

all-MiniLM-L6-v2 is a compact sentence-transformer embedding model for semantic search, retrieval, clustering, similarity scoring, and text ranking.

Space Bunny Alpha

stealth/space-bunny-alpha

Space Bunny Alpha is an anonymous large model with fast inference, strong coding capabilities, native multimodal input, adjustable reasoning effort, and a one-million-token context window. It is operated by an undisclosed third-party provider during this preview.

Union Alpha

stealth/union-alpha

Union Alpha was an anonymous test alias for Unbiased's Pareto 26.9. The alias was retired from provider routing on 2026-09-17; this historical record is retained for documentation and is not enabled for routing.

Ox Alpha

stealth/ox-alpha

Ox Alpha was an anonymous stealth-preview reasoning model for coding, sustained agentic work, and production workloads. OpenRouter reported that the model was revealed as Z.AI GLM-5.3-Flash and is no longer available under the Ox Alpha name.

Uncensored and creative writing model based on Mistral Small 3.2 24B with good recall, prompt adherence, and intelligence.

Skyfall 36B v2 is an enhanced iteration of Mistral Small 2501, specifically fine-tuned for improved creativity, nuanced writing, role-playing, and coherent storytelling.

UnslopNemo v4.1 is the latest addition from the creator of Rocinante, designed for adventure writing and role-play scenarios.

Rocinante 12B is designed for engaging storytelling and rich prose. Early testers have reported: - Expanded vocabulary with unique and expressive word choices - Enhanced creativity for vivid narratives -...

Jev 1.14

typesafe/jev-1.14.0

Jev 1.14 evaluates text, structured records, and images through Phaseo's Decisions API, returning **probabilities, choices, scores, and bounding boxes**. ### Images Send PNG or JPEG base64 data URLs using standard `input_image` parts. Phaseo enables image support automatically; remote URLs are unsupported. Images must be at most **16 MiB**, 256–1568 pixels per side, and at least 78,400 pixels in total. ### Box Use `type: "box"` with instructions naming the target. Each answer returns one rectangle with normalized **0–1 coordinates** and confidence. Box always returns a rectangle, including for absent targets, so pair it with a presence `predicate` and validate accuracy on your images. ### Long context Defaults are **64k tokens per request** and **32k per question**, including shared evidence. Set `provider_options.typesafe.long_context` to `true` to raise both to **262k**. Longer requests can be slower and less accurate. Images, Box, and long context are **experimental**. All input tokens cost **$0.042 per million**, including images and long context; output is free.

Jev 1.13

typesafe/jev-1.13.0

TypeSafe Jev evaluates typed Noul, Choice, and Score questions against structured state.

Pareto 26.9

unbiased/pareto

Pareto 26.9 is Unbiased's multimodal blended model, built and maintained by Circuit & Chisel. It combines multiple frontier and open models in parallel and synthesizes one response. This record is catalogue-only: direct Unbiased access is limited/approval-only and Phaseo has no executable provider route.

A recreation trial of the original MythoMax-L2-B13 but with updated models. #merge

Palmyra X5 is Writer's most advanced model, purpose-built for building and scaling AI agents across the enterprise. It delivers industry-leading speed and efficiency on context windows up to 1 million...